Project Summary Up to 85% of stroke survivors have hemiparesis that affects the upper extremity (UE) on one side and usually impacts the hand more than shoulder and elbow. Currently, for mildly impaired stroke survivors (about 20-25%), constraint-induced movement therapy (CIMT) and modified CIMT have been reporting positive results. However, intervention options for a large percentage of stroke survivors, who have moderate to severe impairment, are lacking. Device use has been studied to assist arm/hand function for individuals with moderate to severe stroke. Positive clinical outcomes have been reported, but the quality of the evidence is low. One of the factors that impact the effects of device-assisted interventions is how the device is used. We suggest that devices should assist the practice of Activities of Daily Living (ADLs) in a way that enhances the neural activities related to ?normal? motor patterns, and minimizes undesired irrelevant activities. We call this ?training ADLs in a close-to-normal pattern?. The importance of practicing ADLs has been demonstrated by previous hand/arm interventions in mildly impaired individuals. When success in ADL tasks becomes the primary goal, individuals usually develop compensatory movements and evoke neural activities unrelated to the required movements. As demonstrated in animal models, compensatory neural activities negatively impact neuroplasticity and motor recovery, while close-to- normal training heightened ipsilesional plasticity and enhanced recovery. This has prompted the opinion that interventions should focus on maximizing motor recovery versus task accomplishment via compensation. We aim to investigate the feasibility of minimizing compensation and maximizing motor recovery in the more severely impaired chronic post-stroke population. Specifically, we propose to use devices to address 2 issues that are commonly presented in this population: 1) inability to open the paretic hand, and 2) abnormal UE synergic movement patterns, defined as the abnormal coupling between shoulder, elbow, wrist, and fingers. Recently, we developed and tested an EMG-triggered functional electrical stimulator (ReIn-Hand) to assist voluntary hand use during the practice of ADLs, and found promising preliminary results in gaining finger extension ability and UE motor function. We also have evidence demonstrating that ACT3D robotic modulation of shoulder abduction loading during actively reaching can reduce the UE synergy both acutely and long-term. By combining ReIn-Hand with an ACT3D robot, we propose a reaching-grasping-retrieving-releasing (GR3) intervention in individuals with moderate to severe chronic stroke. This design aims at practicing ADLs with a ?close-to-normal? movement pattern to achieve functional goals while maximizing potential motor recovery. We will measure not only the intervention-induced changes in clinical outcomes, but also in UE kinematics and functional and morphologic neuroplasticity to disentangle motor compensation versus recovery. If successful, the expected results may impact current clinical practice by pushing towards implementing device-assisted practice of ADLs in a close-to-normal pattern and have the potential to benefit a large population.